Meta Platforms postponed the release of its next-generation large language model, "Behemoth," as internal concerns mount over the model's capabilities and the direction of its multibillion-dollar artificial intelligence investments, The Wall Street Journal reported on Thursday.
Originally planned for release in April to align with Meta's first AI developer conference, the model's rollout was first postponed to June. That timeline has now slipped to fall or even later, according to the report, despite Meta releasing two smaller models from its Llama AI family in the interim.
Engineers working on Behemoth are reportedly struggling to deliver meaningful improvements over previous iterations, leading to questions among staff about whether the model's performance warrants a public launch.
